OpenMagpie is a self-hosted, open-source social listening tool for finding relevant conversations while threads are active. Users describe what they want to find in natural language, and a language model scores new posts against that description. The site lists Reddit, Hacker News, and RSS or Atom feeds as sources available today; plan details also list X/Twitter and YouTube, with historical data described as limited. Matches can be delivered to a webhook or logs, and polls, model judgements, and deliveries are recorded for inspection. Users provide their own hardware and language model. OpenMagpie works with OpenAI-compatible endpoints, including Ollama, vLLM, llama.cpp, LM Studio, and OpenAI. Setup requires Docker and uv, and the current interface is a CLI; a web UI is described as on the way. It is free to run under the Apache 2.0 license. Anonymous usage telemetry is opt-in, off by default, and does not send user content. A hosted version is in development, with an early-access waitlist. Slack and LinkedIn are described as roadmap sources; Facebook, TikTok, and Instagram are planned additions in the repository.
Who it is for
OpenMagpie suits teams or individuals who can self-host a CLI tool, provide hardware and an LLM, and want to monitor relevant conversations. Its current source list includes Reddit, Hacker News, and RSS or Atom feeds.
What is good
- Natural-language descriptions guide post matching.
- Matches can go to a webhook or logs.
- Polls, judgements, and deliveries are recorded for inspection.
- Free to run under Apache 2.0.
What to know first
- Setup requires Docker and uv.
- A web UI is not yet available.
- Users must provide hardware and a model.
- Historical data is limited.
Verdict
OpenMagpie offers self-hosted social listening with user-selected model endpoints and inspectable activity records. Its current CLI setup and the need to provide infrastructure and a model are key considerations.
OpenMagpie plans and pricing
All plansCompared on AI social listening tools
- Free plan
- Yes
- Source coverage
- X/Twitter, YouTube, Reddit, Hacker News, RSS/Atom feeds
- Historical data
- limited
